Post-GWAS Analysis Pipeline

Copyright holder: EMBL-European Bioinformatics Institute (Apache 2 License)

This script is designed to automatically finemap and highlight the causal variants behind GWAS results by cross-examining GWAS, population genetic, epigenetic and cis-regulatory datasets.

Its original design was based on STOPGAP. It takes as input a disease identifier, extracts associated SNPs via GWAS databases, expands them by LD, then searches an array of regulatory and cis-regulatory databases for gene associations.

Installation

Virtual machine

If you wish to shortcut all of the instructions below, you can simply use our VirtualBox virtual machine.

Installing the Python library

Add the lib/ directory to your $PYTHONPATH environment variable.

Installing dependencies

The scripts/installation/ubuntu_environment.sh describes a recipe to install all basic C and Python dependencies on a fresh ubuntu server (requires root access).

To install all binformatic dependencies run sh scripts/installation/install_dependencies.sh.

Add the ./bin/ directory to your $PATH environment variable.

Flatfile preparation

Via the FTP site (recommended)

The following script downloads a bunch of files into $PWD:

  sh scripts/installation/download.sh

Ideally, save these files in a separate directory, which we will call databases_dir.

Manually (sloooow)

The following will create a databases_dir directory for you:

  cd scripts/build_data_files
  make download
  make process

Warning this may take days as it needs to split the entire 1000 Genomes files by population.

Running

Every time you run POSTGAP, add --database_dir /path/to/databases_dir to the command line, where the database directory path corresponds to the directory created above.

By default, run from the root directory the command:

python POSTGAP.py --disease autism --population EUR 

Multiple disease names can be provided.

You can also provide a list of EFOs:

python POSTGAP.py --efos EFO_0000196

Or an rsID:

python POSTGAP.py --rsID rs10009124

Or a manually defined variant:

python POSTGAP.py --coords my_variant 1 1234567 

Analysing your own summary statistics

To short cut the GWAS databases and enter you own data with a file:

python POSTGAP.py --summary_stats tests/sample_data/example.tsv

The summary statistics file should be tab delimited and follow the GWAS Catalog recommentations.

In particular, it must have the following columns:

• variant_id
• p-value
• beta

Bayesian mode (EXPERIMENTAL)

For an EFO, you can trigger the Bayesian calculations with:

python POSTGAP.py --efos EFO_0000196 --bayesian --output2 output2.txt

In this case, POSTGAP produces a tab-delimited output file, 'output2.txt'. The columns represent:

1. Gene ID
2. Cluster description
3. SNP ID
4. Colocalisation posterior probability at that SNP
5. Tissue
6. Colocalisation posterior probability over the whole cluster

It can be displayed as:

python scripts/present_results/postgap_html_report.py --result_file output2.txt --template scripts/present_results/geneReport.html --output report.html

Output

By default, the script writes out a tab delimited file to standard out.

If you wish, you can redirect this into a file:

python POSTGAP.py --disease autism --output results.txt

If you want a JSON dump of all the data retrieved by the pipeline:

python POSTGAP.py --disease autism --output results.json --json
python POSTGAP.py --disease autism --json

Testing

You can check the output with the following commands using the data tests.

More Info

Check out our Wiki